Atlas

Compare Models

Gemini 3.8 Flash vs GPT-5.5

Gemini 3.8 Flash leads by 0.7 capability points (≈95% interval −1.4 to +2.8): statistically tied · P(higher) ≈ 73%

Wins 106 of 197 shared benchmarks (86 lost, 5 tied)

Capability vs. relative cost
Gemini 3.8 FlashGPT-5.5Closed weightsOpen weights

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting.

Gemini 3.8 FlashvsGPT-5.5capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins106 – 86 · 197 shared, 5 tied

Comparing matching setups only (75 benchmarks excluded) · suites counted once
← Gemini 3.8 FlashGPT-5.5 →

Gemini 3.8 FlashvsGPT-5.5wins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 47 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-5.5 minus Gemini 3.8 Flash.

FrontierMath Tier 4 (v2) - PrivateMath22.072.5+50.5
BrokenArXiv 06/2026Math20.869.4+48.6
Olam Social Poker EloGames14791521+45.7
BrokenArXiv 04/2026Math33.672.1+38.5
HieroglyphBenchMultimodal54.622.8−31.8
MASKSafety58.690.1+31.5
BrokenArXiv 05/2026Math19.050.0+31.0
Antidote: Everyday EditionSafety10871013−26.4
LLM Creative Story WritingChat & Writing0.12.4+25.8
ScBench V1 Clean RerunsScience34.057.1+23.1
BrowseComp Extended (Mercor held-out)Search41.519.0−22.5
HardGeoBench Country AccuracyMultimodal67.545.0−22.5
CadQueryEval (Corrected Geometry)Code96.076.0−20.0
HardGeoBench <=100miMultimodal60.040.0−20.0
ArXivMath 05/2026Math58.377.5+19.2
Finance Agent v2 - Market AnalysisFinance & Accounting78.560.3−18.2
TaxCalcBench TY25Finance & Accounting6.024.0+18.0
HardGeoBench <=10miMultimodal46.328.8−17.5
FrontierMath Tiers 1-3 (v2) - PrivateMath68.485.3+16.8
SAGEMath35.151.5+16.5
ArXivMath 06/2026Math67.483.6+16.3
Hemingway-benchChat & Writing10931035−16.1
HPCT Refusal (CAIS)Safety13.029.0+16.0
SM BenchSafety80.465.4−15.0
AA-LCR Extended (Mercor held-out)Search55.670.5+14.9
Vending-Bench 2Games50947524+14.7
SWE-bench Verified Extended (Mercor held-out)Code54.039.3−14.7
Machiavelli Deception (CAIS)Safety86.581.7+14.5
Finance Agent v2 - PrecedentsFinance & Accounting36.121.9−14.2
NoCoT-Bench NCRI 15.2General QA126100−14.0
Roboflow Vision Evals (October 2026, five tasks)Multimodal79.866.0−13.8
AutomationBench-AA 1.0.6Agents60.947.3−13.6
SimpleBenchGeneral QA82.469.0−13.4
TxBench Antibody Discovery V1 Clean RerunsScience43.329.9−13.4
AutomationBench 1.0.6 (Zapier)Agents29.716.6−13.1
Furniture Assembly BenchmarkMultimodal31.744.2+12.5
MindCube TinyMultimodal92.279.9−12.3
Agent Red Teaming k=100 (CAIS)Safety29.841.5−11.7
Surface Evolver BenchAgents76.988.1+11.3
HardGeoBench Avg DistanceMultimodal8431196−10.9
HANDBOOK.mdProfessional Work10.821.5+10.7
Terminal-Bench 4.0 (Artificial Analysis)Code19.79.1−10.6
ChessBench EloGames15271349−10.6
BioTIER RefusalSafety83.973.4−10.5
Time Horizon Index: KSPAgents18.88.3−10.5
Roboflow Vision Evals (September 2026 protocol)Multimodal85.174.8−10.3
TextQuests Average Cumulative HarmSafety24.421.4+10.2
Finance Agent v2 - All PassFinance & Accounting49.739.6−10.1
ARC-AGI-3 (Semi-Private)General QA10.40.4−9.9
ChartographyMultimodal40.931.0−9.9
Finance Agent v2 - Financial ModelingFinance & Accounting32.622.9−9.7
VCT-Refusal (CAIS)Safety6.416.0+9.6
Finance Agent v2 - Disclosure AnalysisFinance & Accounting68.558.9−9.6
DeepSWE v1.1Code73.864.4−9.4
ERQAMultimodal79.670.5−9.1
Finance Agent v2 - AdjustmentsFinance & Accounting54.845.7−9.1
Mystery Game PuzzlesGames47.056.0+9.0
SkillsBench without SkillsAgents41.049.9+8.8
Vibe Code Bench v1.1Code78.769.8−8.8
Code MigrationCode36.545.2+8.6
Finance Agent v2 - ComparablesFinance & Accounting52.043.7−8.4
SpeciEval (Pooled English Runs)Safety89.998.2+8.3
SpeciEval (Refusal-Aware Retries, Historical)Safety89.998.2+8.3
Terminal-Bench 2.1Code87.679.4−8.2
Finance Agent v2 - Earnings AnalysisFinance & Accounting81.473.2−8.2
MBCT (SecureBio)Science48.957.1+8.2
Code2Video-AA EloCode8991000+8.2
SWE-bench MultilingualCode85.877.7−8.1
MedXpertQA MM (200-question set)Health & Medicine84.776.7−8.0
Excel Modeling BenchmarkFinance & Accounting72.264.5−7.7
BioSecBench-Function V1 Clean RerunsScience36.028.3−7.7
MMMU-Pro Extended (Mercor held-out)Multimodal36.443.8+7.4
EnterpriseBench: CoreCraftProfessional Work58.551.3−7.2
CritPt (avg@5)Science18.325.4+7.1
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting80.973.7−7.1
Design Arena Overall Code CategoriesCode12991257−7.0
EpiBench V1 Clean RerunsScience21.428.3+6.9
NoCoT-Bench NCKI (kspine v3)General QA121111−6.9
𝜏³-BankingAgents45.839.0−6.8
ITBench-AAAgents52.545.8−6.7
SimpleQA Verified (Epoch anti-abstention)General QA69.763.0−6.7
AA-Briefcase EloAgents12021089−6.6
FrontierFinance v3.1Finance & Accounting50.043.5−6.5
Terminal-Bench 2.1 Extended (Mercor held-out)Code44.438.0−6.4
Tax Agent BenchFinance & Accounting66.860.5−6.3
BuseyBench SVGMultimodal6.66.0−6.3
PCT Political Inconsistency (CAIS)Safety46.340.0+6.3
Harvey's Legal Agent BenchmarkLegal10.03.8−6.2
LAB-Bench ProtocolQA (SecureBio)Science84.077.9−6.1
GDPval Extended (Mercor held-out rubric score)Professional Work60.454.5−5.9
SUPERChemScience72.666.8−5.8
ChessBench CoherenceGames94.6100.0+5.4
VCT (SecureBio)Science46.952.3+5.4
BrowseComp (Mercor 130-question subset)Search83.177.7−5.4
TextQuests Game ProgressGames54.148.9−5.2
LMArena Agent Arena Bash RecoveryAgents-0.79.5+5.1
Creative Writing v3 EloChat & Writing17441845+5.1
Chess Puzzles (task version 2.0)Games58.063.0+5.0
Vals Index v2Indexes62.357.4−4.8
GDPval-AA v2.1Professional Work14421335−4.7
SpatialBench V1 Clean RerunsScience59.464.1+4.6
SWE Atlas - Codebase QnACode40.645.2+4.6
ArXivMath 04/2026Math64.268.8+4.6
GDPval-AA v2Professional Work14641372−4.5
AA-Omniscience IndexIndexes29.620.5−4.5
HPCT (SecureBio)Science62.466.8+4.4
Public Benefits Bench v1.1Professional Work65.360.9−4.4
SkillsBench with SkillsAgents58.062.2+4.2
LiveCodeBench (Vals AI public dataset)Code89.585.3−4.2
ARC-AGI-2General QA89.285.0−4.2
GameBench 2Games64.059.9−4.1
RiemannBenchMath51.255.2+4.0
SpatialViz-BenchMultimodal78.174.2−3.9
Medical Long Context Reasoning (MLCR-AA)Health & Medicine21.717.8−3.9
CharXiv Extended (Mercor held-out)Multimodal78.982.7+3.8
AA-AnalystAgent (pass^5)Professional Work46.350.0+3.8
GBENCH One-shot Coding GScoreGames58.662.2+3.6
SpatialBench-Long V1 Clean RerunsScience28.632.1+3.6
APEX-AccountingFinance & Accounting51.748.2−3.5
ARC-AGI-1 Semi-PrivateGeneral QA98.595.0−3.5
MortgageTax (Vals all 1,258 documents)Finance & Accounting65.368.8+3.4
MMMU ProMultimodal84.581.2−3.4
AIME 2025Math96.7100.0+3.3
Bio Capabilities Index (SecureBio)Indexes153158+3.1
GeneBench-Pro Public Release SubsetScience11.08.0−3.0
BioSecBench-Surveillance V1 Clean RerunsScience44.147.1+3.0
LMArena Text Arena EloChat & Writing14971481−2.8
RedlineBenchLegal41.344.1+2.8
GDP.pdfProfessional Work23.226.0+2.8
ForecastBench Preliminary Dataset Brier IndexForecasting60.563.3+2.8
LAB-Bench CloningScenarios (SecureBio)Science100.097.3−2.7
APEX-Agents 1.1Professional Work72.970.3−2.6
IntPhys 2Multimodal62.559.9−2.6
SWE-bench VerifiedCode80.082.6+2.6
WCB (SecureBio)Science57.159.7+2.6
ALE-BenchCode21822108−2.6
Artificial Analysis Intelligence Index v4.3.2Indexes40.938.4−2.6
Artificial Analysis Intelligence Index v4.3Indexes41.238.6−2.6
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting68.165.6−2.5
Vision Arena OverallMultimodal12931283−2.5
ChessBench AccuracyGames81.178.8−2.5
MineBench — Bradley-Terry RatingGames18061846+2.4
Blueprint-Bench 2Multimodal0.40.4−2.4
Humanity's Last ExamGeneral QA46.243.8−2.4
MedScribeHealth & Medicine84.586.9+2.4
ACCoRDGeneral QA78.280.5+2.4
CRM-Agent-BenchAgents83.681.3−2.3
Conceptual Reasoning IndexIndexes65.767.3+2.3
Document ArenaProfessional Work14821484+2.1
SciCode (main problems)Science14.212.3−1.9
LLM Debate BenchmarkChat & Writing15141528+1.9
HiL-Bench (Human-in-Loop Benchmark)Agents41.539.7−1.8
MMLU ProGeneral QA90.788.9−1.8
SpatialBench V0 VerifiedScience67.065.2−1.7
LAB-Bench LitQA2 (SecureBio)Science51.549.8−1.7
ScBench-Long V1 Clean RerunsScience19.921.6+1.7
LMArena Agent Arena Net ImprovementAgents3.84.2+1.6
PG-LLM-50Science0.40.3−1.5
Creative Writing v3 Rubric ScoreChat & Writing15.615.9+1.5
Legal Research BenchLegal38.940.4+1.4
Longform Creative WritingChat & Writing76.878.2+1.4
LMCAGeneral QA52.954.3+1.4
LiveBench 2.0Agents67.368.7+1.4
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine13.314.7+1.3
LMArena Agent Arena SteerabilityAgents4.36.9+1.3
Humanity's Last Exam DiamondGeneral QA29.127.9−1.2
Extended NYT Connections 940 (Quadratic v1)Games97.496.2−1.2
Epoch Capabilities Index (ECI)Indexes159160+1.1
TxBench Oligo Discovery V1 Clean RerunsScience33.131.9−1.1
OTIS Mock AIME 2024-2025Math98.9100.0+1.1
ComplexConstraintsChat & Writing48.449.5+1.1
GDPval (Mercor 213-task rubric score)Professional Work83.584.6+1.1
DAXBench v2Code87.288.3+1.1
SEA-HELMGeneral QA77.376.3−1.0
WordleBenchGames98.097.0−1.0
MedCodeHealth & Medicine48.149.1+1.0
VariantBench V1 Clean RerunsScience31.430.5−0.8
CharXiv (descriptive and reasoning combined)Multimodal95.494.6−0.8
TxBench Preclinical Pharmacology V1 Clean RerunsScience54.755.3+0.7
LLM Chess EloGames15461532−0.6
GPQA DiamondGeneral QA94.995.5+0.6
TaxEval v2Finance & Accounting74.475.0+0.5
ProgramBenchCode1.00.5−0.5
LegalBenchLegal87.086.5−0.5
SciCode 1.0.1 (Artificial Analysis)Code56.656.1−0.5
APEX-1 (Mercor, mean score)Professional Work68.868.4−0.4
AA-LCR v1.1Search84.084.3+0.3
DTBenchGeneral QA95.796.0+0.3
APEX-SWECode36.436.6+0.2
WeirdML (v2)Science84.884.9+0.1
Artificial Analysis Progress Index v1Indexes320320−0.1
VCT v2 (SecureBio)Science45.245.3+0.1
BabaIsBench Stage 0: The IntroGames100.0100.00.0
EnigmaEvalGeneral QA37.237.20.0
FrontierMath ErdősMath0.00.00.0
MedXpertQA MM Extended (Mercor held-out)Health & Medicine60.860.80.0
ObviousBenchGeneral QA100.0100.00.0

Gemini 3.8 FlashvsGPT-5.5normalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 47 othersVillage LLM Atlasatlas.village.dev

Win counts use each benchmark’s declared score direction. Represented suite slices and the derived Capability score are excluded from headline totals. Missing or incompatible setups are excluded, never counted as losses.
Loading Atlas data…